While beta from CAPM only considers market returns, news beta tries to capture the underlying factors that drive prices typically reported in financial news.
This study shows that positive news beta stocks on average outperform the market, while negative news beta stocks tend to underperform.
Beyond using news beta as an additional risk measure, I find that in a long/short trading strategy, going long positive news beta stocks and short negative news beta stocks yields a positive return spread over a 10 year back-testing period.
Finally, I show how it is possible to improve a regime-shift detection strategy by tracking the ratio of negative to positive news beta stocks. Trading the S&P 500 index either long or short based on such strategies, I obtain an information ratio of approximately 1.0 with annualized returns of 16 percent and a monthly hit ratio above 65 percent.
1. Introduction: News Beta
Improving the ability to predict stock prices is one of the central goals of investors for all of the obvious reasons. Several factors including fundamentals and sentiment are likely to impact the price formation process. Identifying valuable sources of information is essential to stay at the cutting-edge in the alpha generation process. Financial news has always been considered an important source of information, but only in recent years have investors been able to systematically include news in a trading strategy and in a quantitative manner.
For example, Tetlock finds that ten-day reversals are reduced following company specific news, which indicates that public news is a proxy for information that has not yet been incorporated into the stock price [Tetlock, 2009]. Engelberg et al. find that short seller trades are more than twice as profitable in the presence of recent news, providing evidence that news can yield profitable trading opportunities for skilled information processors [Engelberg et al., 2010]. Mitra et al. included news sentiment as part of the construction of forward-looking covariance matrices finding that sentiment could add value to the volatility prediction process beyond what could be captured by option-implied volatility [Mitra et al., 2009]. Also, Zhang and Skiena have shown that news is significantly correlated with both trading volume and stock returns [Zhang & Skiena, 2010].
Cahan et al. not only found that the impact of scheduled and unscheduled news events can be measured in days, weeks, and months; but also that quant factors based on news analytics could add value to traditional multifactor models [Cahan et al., 2009a, Cahan et al., 2009b]. More recently, Conomos et al. found that using news sentiment as an overlay to a momentum strategy significantly improved the annualized return spread from 20.1% to 31.8%, as well as increased the Sharpe ratio by about 50 percent [Conomos et al., 2010].
Using a one year investment horizon, Kittrell found value in detecting net sentiment reversals as a measure for long-term stock selection. More specifically, he found that such signals on average outperformed the S&P 500 by 16 percent per year from 2000 through 2008, both in bull and bear markets [Kittrell, 2010]. In previous research I found that a long/short strategy trading the top and bottom 5 ranked industries would have yielded an information ratio (IR) of 1.23 over a five year back-testing period [Hafez, 2010].
By incorporating structured news data or news analytics in a trading model, investors can react to scheduled and unscheduled news events in real-time and in a semi- or fully automated manner. Representing news as a structured and quantitative metric also means one can measure the prevailing sentiment trend on a given market. Such trends can be captured by looking at aggregated news sentiment on single companies, sectors, or even on broader equity portfolios. Capturing sentiment at a market level is important, since many experts are of the opinion that the entire market dictates approximately 70 percent of an individual stock’s movement whether up or down, while company-specific news or actual performance accounts explain the remaining 30 percent [Endoo8, 2010].
In finance, beta describes the relationship between a company’s stock return with that of a market benchmark. A zero beta means that the price of a stock is uncorrelated with the market, while a positive or negative beta stock means that the price generally follows or inversely follows the price of the market, respectively. Beta can be considered a normalized covariance measure in terms of units of market return. Hence, if the market benchmark moves by one unit, the stock price is expected to move by beta units. The beta coefficient is a key parameter in the Capital Asset Pricing Model (CAPM). It measures the part of an asset’s statistical variance that cannot be mitigated by diversification (also known as systematic risk). In practice, beta can be estimated for individual stocks using regression analysis against a stock market index.
Traditionally, if an investor is evaluating a given stock for his or her portfolio and wants to gauge its potential volatility in relation to the market as a whole, understanding the market beta is invaluable. Interestingly, Patton and Verardo introduce news into the measurement of market beta and find that it increases by an economically and statistically significant amount on days of company-specific news announcements, after which it reverts to its average level two to five days later [Patton & Verardo, 2009]. Having previously shown that market sentiment can be a valuable input when predicting market direction [Hafez, 2009], in this paper I consider the responsiveness of individual stocks to such market sentiment measure. With this objective, it seems natural to echo the market beta approach and introduce the concept of a (market) news beta, which is to be considered one in a series of tools to incorporate news into an investment strategy.
Similar to the beta in CAPM, news beta measures the responsiveness of a company’s stock price to a market benchmark. While beta from CAPM only considers market returns, news beta tries to capture the underlying factors that drive prices as reported in the media. This is done by calculating a market sentiment index. In order to capture the potential dynamic relationship between market sentiment and stock returns, I propose that news beta be estimated based on a rolling regression model...
Click here to download the PDF and continue reading the "News Beta - A New Measure for Risk and Stock Analysis" White PaperRequest White Paper